Brain (1983), 106, 623-642

TIME OF CONSCIOUS INTENTION TO ACT

IN RELATION TO ONSET OF CEREBRAL
ACTIVITY (READINESS-POTENTIAL)
THE UNCONSCIOUS INITIATION OF A FREELY

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

VOLUNTARY ACT

by BENJAMIN LIBET, CURTIS A. GLEASON, ELWOOD W. WRIGHT and

DENNIS K. PEARL1
{From the Neurological Institute, Department of Neuroscience, Mount Zion Hospital and Medical
Center, the Department of Physiology, School of Medicine, University of California, San Francisco,
CA 94143 and the Department of Statistics, University of California, Berkeley, CA)

SUMMARY
The recordable cerebral activity (readiness-potential, RP) that precedes a freely voluntary, fully
endogenous motor act was directly compared with the reportable time (W) for appearance of the
subjective experience of'wanting' or intending to act. The onset of cerebral activity clearly preceded by
at least several hundred milliseconds the reported time of conscious intention to act. This relationship
held even for those series (with 'type II' RPs) in which subjects reported that all of the 40 self-initiated
movements in the series appeared 'spontaneously' and capriciously.
Data were obtained in at least 6 different experimental sessions with each of 5 subjects. In series with
type II RPs, onset of the main negative shift in each RP preceded the corresponding mean W value by
an average of about 350 ms, and by a minimum of about 150 ms. In series with type I RPs, in which an
experience of preplanning occurred in some of the 40 self-initiated acts, onset of RP preceded W by an
average of about 800 ms (or by 500 ms, taking onset of RP at 90 per cent of its area).
Reports of W time depended upon the subject's recall of the spatial 'clock-position' of a revolving
spot at the time of his initial awareness of wanting or intending to move. Two different modes of recall
produced similar values. Subjects distinguished awareness of wanting to move (W) from awareness of
actually moving (M). W times were consistently and substantially negative to, in advance of, mean
times reported for M and also those for S, the sensation elicited by a task-related skin stimulus
delivered at irregular times that were unknown to the subject.
It is concluded that cerebral initiation of a spontaneous, freely voluntary act can begin
unconsciously, that is, before there is any (at least recallable) subjective awareness that a 'decision' to
act has already been initiated cerebrally. This introduces certain constraints on the potentiality for
conscious initiation and control of voluntary acts.

1
Present address: Department of Statistics, Ohio State University, Columbus, Ohio.
Reprint requests to Dr B. Libet, Department of Physiology, University of California, San Francisco,
CA 94143, USA.
624 BENJAMIN LIBET AND OTHERS

INTRODUCTION
The 'readiness-potential' (RP), a scalp-recorded slow negative potential shift that
begins up to a second or more before a self-paced act (Kornhuber and Deecke,
1965; Gilden et al., 1966), can also precede self-initiated 'freely' voluntary acts which
are not only fully endogenous but even spontaneously capricious in origin (Libet et
al., 1982). The appearance of preparatory cerebral processes at such surprisingly
long times before a freely voluntary act raises the question of whether conscious
awareness of the voluntary urge or intention to act also appears with such similar
advance timings. The present study attempts to answer this question experimentally.

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

In the present study, the experience of the time of thefirstawareness of the urge to
move was related by the subject to his observed 'clock-position' of a spot of light
revolving in a circle; the subject subsequently recalled and reported this position of
the spot. Thus the experience of timing of the awareness was converted to a
reportable, visually related spatial image, analogous to reading and recalling the
clock-time for any experience. (The reliability and validity of this operational
criterion are further considered below.) This indicator of the time of the conscious
experience could then be related (1) to the actual time of the voluntary motor act, as
indicated by the electromyogram (EMG) recorded from the appropriate muscle,
and (2) to the time of appearance of the simultaneously recorded RP that is
generated by the brain in advance of each act. The voluntary motor acts under study
were those produced with minimal or no restrictions on the subject's independent
choice of when to act, and under instructions that encouraged spontaneity of each
volitional urge to act (Libet et al., 1982).
The present findings thus provide experimental evidence on the timing of the
conscious intention to act relative to the onset of cerebral activity preparatory to the
act, and on the roles of conscious processes in the initiation of a freely voluntary
motor act.

METHODS AND PROCEDURES

Subjects
Six right-handed college students were studied as two separate groups of three each. Group 1
comprised 3 females (S.S., CM. and M.B.), but the quality of the EEGs and the minimal amplitude of
the RPs of one precluded using much of her data. Group 2 consisted of 2 males and 1 female (S.B., B.D.
and G.L.). Study of this group began a few months after completing the study of Group 1.
Recording
The d.c. recording and averaging of the EEG has been described (Libet et al., 1982). For present
purposes, analysis of RPs is made for those recorded at the vertex, where they were all maximal. (For
the first 4 experimental sessions with Group 1, only the contralateral precentral recording site is
available.) Linked mastoid electrodes served as the reference lead, with a ground electrode on the left
ear lobe. Controls excluded the electro-oculogram as a source of the slow potentials. In each
experimental series, 40 trials were performed and averaged by a computer of average transients (CAT
400B). The 2 s period of EEG stored by the CAT with each trial included a 1.4 s period already on the
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 625

recording tape before 'O-time'. The latter was signified by the EMG, recorded with bipolar electrodes
on the skin over the activated muscle of the right forearm.

Procedure
The subject sat in a partially reclining position on a lounge chair with an observer present in the
room. Each trial was started only when the subject considered himself comfortably ready. The trial
began with a brief'get-ready' tone. This signalled that during the next 1 -3 s the subject should relax his
muscles, especially those of the head, neck and forearm, blink his eyelids if he wished, and fix his gaze
on the centre of the 5 inch circular screen of a cathode ray oscilloscope (CRO) that was positioned at
about 1.95 m away in his direct line of vision. At the end of these irregular get-ready periods the
operator activated the PDP-12 computer to initiate circular revolution of the beam of the CRO. The
CRO spot of light revolved in a clockwise circle near the circumference of the screen starting from

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

the' 12-o'clock' position; this motion simulated a sweep secondhand of a clock but each revolution was
completed in 2.56 rather than 60 s. A circular scale, with numbers at each '5 s' position, was mounted at
the external edge of the CRO screen, and a plastic grille on the peripheral portion of the screen
displayed illuminated radial lines spaced at '2.5s.' intervals (each equal to 107 ms of actual time).
Subjects were asked to maintain their gaze fixed on the centre of the CRO screen and not to follow the
CRO spot around, even though they were to report information relating the 'clock-position' of the spot
to the events (see below). The visual angle subtended between the centre and the peripheral position of
the moving spot was small enough (1.8 deg) to present no difficulty from loss of visual acuity. The
'clock-time' of the CRO spot at each event, namely EMG with motor act or stimulator synch pulse with
stimulus to skin, was recorded by the PDP-12 computer. Subjects were trained to make their self-
initiated movement sufficiently brisk so that within no more than 10-20 ms from the start of any EMG
potentials they achieved the amplitude pre-set to trigger the computer.
The subject was asked not to blink from the time the CRO spot started revolving until after the event.
To minimize the possibility that the need to blink might become a controlling 'external' factor that
compels or impels him to act, the subject was told that he may blink during the trial if the need arose;
but that, if he did blink (or made some other extranaeous motion), he should then simply wait for the
CRO spot to make at least another full revolution before performing the quick voluntary movement, as
at the start of the trial.
Two different kinds of series were studied.
(1) Self-initiated voluntary acts. The subject was asked to wait for one complete revolution of the
CRO spot and then, at any time thereafter when he felt like doing so, to perform the quick, abrupt
flexion of the fingers and/or the wrist of his right hand (see Libet etai, 1982). An additional instruction
to encourage 'spontaneity' of the act was given routinely to subjects in Group 2 and only in the latter
half to two-thirds of sessions with Group 1. For this, the subject was instructed 'to let the urge to act
appear on its own at any time without any preplanning or concentration on when to act', that is, to try
to be 'spontaneous' in deciding when to perform each act; this instruction was designed to elicit
voluntary acts that were freely capricious in origin.
(2) Skin-stimuli 'at unknown times'. For such a series the subject expected to receive a near-threshold
stimulus pulse on the back of the right hand. Delivery of the pulse was made by the operator at irregular
times that were unknown to the subject, but only after the CRO spot completed its first revolution.
They were actually delivered randomly during the second or third revolution of the spot (that is,
between about 2.6 and 7.6 s after the spot began to revolve); this range overlapped with that for the
times of the self-initiated movements. These conditions closely paralleled the attentive and other
requirements associated with performing and recalling the CRO clock time for 'spontaneous' self-
initiated voluntary acts (see also Libet et al., 1982).
Subjects' reports of the time of an event. The 'clock position' of the revolving CRO spot at the time of
the subject's awareness of an event was observed by the subject for later recall. Within a few seconds
after the event, the subject was asked for his report of that timing, as in recalling a spatial image of
ordinary clock time in conjunction with another event. It was emphasized that only an after-the-event
626 BENJAMIN LIBET AND OTHERS

recall of the experience was required, and that the subject should not worry about the task in advance
of each event. Subjects became rapidly accustomed to this task during the training runs and did not
find it to be taxing or stressful; nor did this task have any detectable effect on RPs (Libet et al, 1982).
Modes of recall. Although each report depended on the subject having continously monitored the
revolving CRO spot and visually noting, to himself, the position of the spot at the actual time of his
awareness (of the event under studysee below), two different modes were employed for his after-the-
event recall of that spot position. With the (A) or 'absolute' mode, the subject was asked to look back
on the circular time scale mounted on the CRO and report the 'clock-time' of the spot position in
'seconds'. (Each 'second' on this scale corresponded to an actual time of 2560/60 or about 42.7 ms.)
With the (O) or 'order' mode, the subject was asked to report the order of the final stopped position of
the CRO spot, at the end of the trial, relative to his recalled position of the moving spot at the time of his
awareness. For this, the subjects simply reported 'CRO spot (stop-position) first', at an earlier clock-

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

position than the event-awareness; or 'awareness first', or 'together' (same position for both, insofar as
the subject could discriminate). The (O) mode of recall was found by most subjects to be somewhat less
demanding than the (A) mode.
The final stop position for the CRO spot following each event was arrived at in a complex manner
that differed for (A) or (O) mode of recall. When either of the modes of recall was to be used, the computer
continued the clockwise motion of the spot for a period beyond the time triggered by the event; this was
called the 'contination interval'. The continuation interval could have one of 20 different values, all in
the range between +500 and +800 ms (approximately 12 to 19 'seconds' of clock-dial). One of these
continuation values was selected by the computer, from a randomized series of the 20, for use in a given
trial. No violations of independence of answer were found in relation to the randomized continuation
intervals.
For the (O)-order-mode only, however, the CRO spot did not stop after reaching its continuation
interval; instead the spot jumped discontinuously, to stop at clock positions that were both before and
after the subject's recalled positions for his awareness. (1) 'Stopping range'. The clock-times within
which all of these final stop positions were included (the 'stopping range') ordinarily spanned 600 ms of
real time. The positive and negative end points of the stopping range, relative to zero trigger time for
each event (EMG or S-synch pulse), were chosen so as to span the entire range of times (relative to O) in
the reports for a given awareness (W, M or S, defined below). We usually succeeded in setting the
positive end of the stopping range well beyond the stop-times of the CRO spot for which all reports
were 'W-first' (or M- or S-first), that is, earlier than the stopped spot position; and the negative end of
the stopping range well beyond the stop-times for which all reports were 'spot first' (as in fig. 1B).
Actual distributions varied with the subject. The beginning and end points of the 'stopping range' were
therefore set individually before each series of trials, depending on previous results with the kind of
awareness to be reported and with that subject (see examples in fig. 1); the training series of 10 trials
that routinely preceded each regular series of 40 trials was useful for this purpose. (2) 'Stop-times'.
Within each selected stopping range that spanned 600 ms, one of 40 different actual stopping times for
the CRO spot, at intervals of 15 ms, was randomly selected by the computer for use after each of the 40
events. A different sequence of these randomized stop times could be preselected for the successive 40
trials in given series, so that a given sequence of stop-times was not repeated in a given session. The
length of the 'continuation-interval', that precedes the final jump of the spot to its stop-time (see above),
was randomly varied in a fashion that was independent of the randomized sequences selected for the
final stop-times. The objective of all this was of course to avoid providing any clues that might relate the
stop-time position of the CRO spot to the clock-time of the event itself.
The O procedure would not seem to be subject to the kind of artefactual difficulty described by
Garner (1954). In the latter's case, subjects were asked to judge 'half-loudness' referred to a standard
acoustic stimulus; each subject gave reliably consistent judgements, but these turned out to be
appropriate only with reference to each different range of stimulus intensities presented rather than to
the standard stimulus. In our case, the stopping range of reference times was determined for each
subject from his own range of reported W times, as indicated by initial trials, rather than vice versa.
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 627

There were some series for which the adopted stopping range did not appropriately span the full range
of potential reports by the subject; that is, the numbers of'W first' and 'spot first' reports were far from
equal, with few or no reports for one of these possible responses. This indicates that the subject was not
shaping his reports to correspond to the adopted range of stop-times for the CRO spot. In such
instances, the series had to be repeated, in the same or a later session, with a more suitable stopping
range that could result in a statistically usable analysis for the mean W time in the series. Additionally,
the subject had no prior knowledge of or consistent experience with the actual stopping ranges that
were used. In a usual given experimental session, each separate series for W, M and S awareness times
required a different stopping range. This meant that the subject did not have any consistent stopping
range with which he was repeatedly presented in successive series.
Type of 'awareness' to be timed by the subject. Three different subjective experiences were to be timed.
Each series of 40 trials involved reports of only one of these for all 40 events.

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

(W). In W series, the subject was asked to note and later report the time of appearance of his
conscious awareness of'wanting' to perform a given self-initiated movement. The experience was also
described as an 'urge' or 'intention' or 'decision' to move, though subjects usually settled for the words
'wanting' or 'urge'. Subjects were told there were no right or wrong or inaccurate answers possible in
the W and M series; that we simply wanted their best recollection and report of their own experience.
Subjects were also asked, after completing a series of 40 self-initiated acts, whether any of the acts
occurred without their experiencing any prior awareness of an urge or intention to act, that is, whether
they were 'surprised' to realize a voluntary act had occurred without a conscious prior intention. In
three series, subjects (3 of the 5 studied) did report that some, usually a small number, of the 40 acts
'surprised' them (Libet et al., 1982), although the mean W time for such series of 40 was negative to
EMG zero time, as usual. Asking subjects about 'surprise' acts should have indicated to them that it
was acceptable even to have and to report the absence of a conscious urge or intention to act prior to a
self-initiated act. The fact that instances of 'surprises' were reported increases confidence that the
reports of timing prior to the act represented endogenous experiences not defined or induced by the
instructions.
It was emphasized that the reported time should refer to the earliest appearance of the W awareness
that might precede an actual movement; the subjects easily distinguished this specific urge to act from
any awareness of a more general preplanning nature that might precede W, as in association with type I
RPs (see Libet et al., 1982), and also from an occasional urge to move that was not followed by an
actual movement.
(M). In M series, the time of subject's awareness that he/she 'actually moved', in the self-initiated act,
was to be noted and reported after each trial. This was intended to provide an instruction for timing an
experience related to but different from that of W, and thus to act as a partial test of the validity of the
W timings. Some subjects stated, on their own, that their mental set differed somewhat in W M M
series. In a W series there was a feeling of active attention to or 'watching for' the awareness of wanting
to move, so as to be able to note the time of its appearance, although the urge to move arose
spontaneously with no preplanning; but M series proceeded without such advance 'watching' for the
event to occur. It should be emphasized that any such difference in mental set did not appear to affect
the associated RPs, which had similar forms and onset times for W and M series (see Results, below).
(S). In S series, time of awareness of the sensation elicited by the near-threshold stimulus pulse to the
back of the hand, delivered at randomly irregular times unknown to the subject, was to be noted and
reported after each trial. The attentive and other conditions of this task closely paralleled those for the
W and M series, except that the event was an externally-induced sensory instead of a self-initiated
motor one. After each S series, whether for training or experimental purposes, the subject was given a
rough indication of how close he was to the actual times for the stimuli, as an aid in improving accuracy
in all the experiments. If the awareness times reported in an S series were to differ significantly from the
actual times of the stimulus pulse in each respective trial, the mean difference between the reported and
actual stimulus times may be regarded as a measure of that subject's 'bias', when observing and
reporting awareness times under the experimental conditions employed (see Libet et al., 1979). This
628 BENJAMIN LIBET AND OTHERS

would include errors in making simultaneous judgements of the times for a mental event (sensation
here) and a visual event (position of the CRO spot) (see Discussion). There were in fact mean
differences or 'shifts' for S that were characteristic for each subject.
Progression of the experiments. The first (and, in some cases, the second) half-day session was purely
for training purposes. Subsequently, each subject was studied in 6 to 8 regular half-day sessions,
usually 1 per week. Each of the first 4 regular sessions began with a training series of 25 trials with skin
stimuli intended as a retraining of reporting the recalled times of a subjective experience. This series
differed from the experimental S series (of 40 trials) in that, after each 5 trials of this series, the subject
was told what the actual 'clock-times' were for the skin stimuli, whose randomly irregular delivery
times were not known before the trials. Following this there were two 40-trial series of self-initiated
acts, one W and one M, each preceded by a briefer 10-trial series for retraining purposes, and then one
40-trial series of skin stimuli delivered at unknown times (S). The order of W and M series within a

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

session was alternated for each successive session. A given mode for subjects to recall clock-positions of
the revolving spot, (A) vs (O), was used for all series in a given session, but the modes were alternated
for successive sessions. In sessions after the first 4, few M or S series were studied.

I
+128(ms)
'+2.5 s'

(W) time reported

I
W-lst

Spot-1st IT
-400 -250 -100 50 200 (ms)
Stop position of CRO spot
FIG. 1. Examples of plots of a subject's reports of (W) time after each trial in series of 40 self-initiated, voluntary
movements. A, series in session 3 for subject S.B. mode of recall is 'absolute' (A); i.e. subject reported 'clock-time' of
CRO spot at time offirstawareness of'wanting' to move. From this value, the 'clock-time' of the EMG-trigger was
subtracted; the resultant net (W) time, relative to EMG zero time, was plotted. Abscissa indicates both the net 'clock-
time' in seconds on the dial (arrows) and the corresponding net real times for (W). B, series in session 2 for subject
S.B. mode of recall is by 'order' (O). The subject reported one of three alternatives for his recalled position of the
revolving spot at the initial awareness of W. As indicated on the ordinate these were 'W-first' (W time earlier than
thefinalstop-position of the CRO spot); or 'Spot-first'; or 'together', T (W time indistinguishable from or same as
for thefinalstop-position of the spot). The 'stopping range', within which lay all the 40 different stop-positions of
the spot, randomly sequenced in the series of 40 trials, was set between 200 ms positive to (i.e., later than) and 400 ms
negative to the EMG zero time in each trial. (Since both ends of the 600 ms stopping range were included, the
computer actually designated 41 stop-positions at 15 ms intervals, but the subject did not report W time for the 41 st
trial of the series. This accounts for an absence of a report at the stop-time of 115 ms in thefigure,which happened
to be the randomly sequenced stop-time of the CRO-spot in the 41st trial of this series.)
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 629

Statistical handling of the response times. This is described in relation to the data in two actual series,
one in (A) and one in (O) mode of recall (plotted in fig. 1). For each event in a series of 40 with (A)
mode of recall, the 'clock-time' in 'seconds' for EMG-trigger time is subtracted from the 'clock-time' of
the CRO spot at the time of awareness, as reported by the subject; this gives a net reported clock-time
(relative to 'EMG zero time'). Each net reported clock-time is then converted to real time, and these W
times (reported real time of each awareness of wanting to move, relative to EMG zero time) are plotted
for that series. (Each 60 s of 'clock-time' = 2560 ms actual time.) For example, in fig. 1A we see that
subject S.B. reported W times of 43 ms twice, O ms 4 times, 43 ms 8 times, etc. (as converted from
net reported clock-times of 1, 0, 1 'seconds', respectively). Averaging these values for the whole
series gives a mean shift for W, relative to EMG-O time, of 2.1 'seconds' of clock-time or 90 ms
of real time.
For the (O) mode of recall we have extended to trinomial data the idea presented by Church and

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

Cobb (1971). With this technique, the mean W shift was calculated as:
(upper, positive end of'stopping range') (time interval between 'stop times') x (number of points 1 /2)
'Stopping range' and 'stop-times' are defined above. 'Number of points' is calculated by giving 1 point
for each time the subject says 'W first' (i.e. spot position at time of awareness has an earlier 'clock-time'
than the final 'stop-time' of the CRO spot), and 1/2 point for each response of 'together' (i.e. W time
and stop-time of spot appear the same to the subject). (In some series a trial was 'aborted' or a subject's
report was not available, for some technical reason. In such a case that trial was considered to
contribute a number of points equal to our estimate of the probability that the subject would say 'W
first' for that particular 'stop-time' of the CRO spot.) In the example shown in fig. 1 B, the responses for
each of the different 'stop-times' are plotted. In that series, the upper end of the 'stopping range' was
200 ms after EMG-trigger time (lower end was 400 ms), with the usual minimum 15 ms time interval
between stop-times within the 'stopping range' of 600 ms. There were 19 'W first' responses and 8
'Together' responses; the remaining responses were of course 'Spot first'. There was one trial with a
missing report in this series, at the CRO stop-time of 115 ms; we estimate the probability of saying
'W first' as 1/2 in this case. Putting this together, our estimate of the mean shift for W is

200-15(19 + 8/2 + 1/2-1/2)= - 1 4 5 ms

The mean shifts for the awareness in an M or S series were computed in an analogous manner for A or
O mode of recall, respectively, using the stimulator-synch trigger for zero time in the S series.

RESULTS

/. Subjective Timings
The mean values of the 40 reported times of awareness (whether for W, M or S),
for each series in a given study session, are presented in Table 1, A. Each value is for
net time relative to 'zero-time' for each event, that is, W or M relative to EMG zero
time for activation of muscle in a self-initiated movement, or S relative to stimulus-
pulse time in the case of skin stimuli delivered at irregular times. There were no
obvious or consistent differences between sessions in which mode of recall of time
was 'absolute' (A) or by order (O) relative to final 'clock-position' of the spot (see
Methods).
The mean value of W in each series, that is of the recalled times for being aware of
'wanting' to move, was invariably in advance of or negative to the EMG zero time.
The average of all such mean Ws was about 200 ms (Table 2D). Except for the
nature of the event, the basic procedures for attentive monitoring of the revolving
630 BENJAMIN LIBET AND OTHERS

spot and of noting visually and later recalling the clock-position of the spot, in
connection with appearance of an awareness, were the same for M ('actually
moved') or for S (skin sensation produced by irregularly timed, stimulus pulse) as
they were for W. Reported times for S might be expected on the average to be close

TABLE 1
Column A. Awareness times (W) and, column B, RP-onset times, for each 'W series' of 40 self-initiated movements.
Awareness times also given for (M) and (S) series in same session. RP onsets are given for both the 'main negative
shift' (MN) and for time at which 90 per cent of total area under the RP begins (see text for definitions of W, M and
S). Column c. Differences (ms) between RP onset-times and W times ('uncorrected', and 'corrected' for S), taking
RP onset either for the main negative (MN) component or for 90 per cent of the RP area, in each W series of 40 self-
initiated acts. (Instances in each series of 40 trials when the W time preceded [was negative to] onset of RP are given

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

after RPMN onset for the respective series.)
A. Awareness times (ms) c. (Onset RP) minus (W),
B. Onset of RP (Wseries) i.e. (B)-(A), using (Onset Rl
f Fninus
Mode 0 (M) (S) (W-S), using
Sub- Ses- Type Kr
MN 90%
ject sion call X SE1 It SE It SE Rf MN- 90 % area toMN RPMN RP
90%
S.B. 1 A - 54 11 - 21 15 - 12 11 II - 550 -1076 - 496 0 -1022 - 508 -1034
2 O -145 19 - 48 20 - 53 15 II - 900 - 729 - 755 0 - 584 - 808 - 637
3 A - 90 11 - 72 12 - 42 12 I -1100 - 863 -1010 0 - 773 -1052 - 815
4 O -188 34 - 95 20 - 53 18 I -1150 - 757 - 977 0 - 584 -1030 - 637
5a A -123 16 II - 800 - 876 - 677 0 - 753
b A -119 10 I - 950 - 694 - 831 0 - 575
6a A -118 13 I - 900 - 685 - 782 0 - 567 - 625 - 410
b A -161 15 +157 II - 600 - 484 - 439 0 - 323 - 282 - 166

G.L. 1 O -208 28 -213 28 -147 28 II - 500 - 380 - 292 0 - 172 - 439 - 319
2 A -422 24 -172 22 -184 30 I -1200 - 755 - 778 0 - 333 - 962 - 517
3 O -377 27 -220 26 -217 30 I - 900 - 635 - 523 0 - 258 - 740 - 475
4 A -258 21 -201 25 -120 25 II - 800 - 593 - 542 0 - 335 - 662 - 455
5a 0 -213 42 I - 900 - 599 - 687 0 - 386
b 0 -283 34 I -1200 - 866 - 917 0 - 603
6a 0 -221 34 I II - 600 - 563 - 379 0 - 342 - 543 - 506
b O -271 40 164 20 I -1400 - 765 -1129 0 - 494 -1293 - 658
1
B.D. 1 A -225 19 + 92 10 + 135 13 II - 400 - 295 - 175 3 - 70 - 40 + 65
2 O -145 24 - 3 25 + 45 26 III - 225 - 157 - 80 1 - 12 - 35 + 33
3 A -152 14 + 76 12 + 61 9 II - 500 - 401 - 348 0 - 249 - 287 - 188
4 0 -142 18 + 40 21 + 90 20 II - 425 - 469 - 283 0 - 327 - 195 - 239
5a O -145 29 III - 250 - 716 - 105 0 - 571
b 0 -108 46 III - 325 - 210 - 217 | - 102
6a 0 -146 30 II - 650 - 468 - 504 0 - 322

S.S. 1 0 -235 31 -168 25 -130 27 Ic - 250 - 806 - 15 3 - 571 - 145 - 691

2 0 -253 28 - 33 19 - 83 17 11 - 400 - 282 - 147 0 - 31 - 230 - 112
3 0 -255 26 -153 24 - 75 24 nc - 400 - 281 - 145 2 - 26 - 220 - 102
4 A -283 19 -113 9 -157 13 mc - 300 - 695 - 17 17 - 412 - 174 - 569
7a A -248 20 v - 900 - 915 - 652 0 - 667
b A -236 17 - 400 - 604 - 164 4 - 368
c A -209 16 i -1100 - 805 - 891 0 - 596

CM. 1 O -287 -138 - 63 - 500 - 408 - 213 - 121 - 274 - 184

(30 trials)
2 0 -223 25 -123 21 - 23 20 "c - 400 - 489 - 177 0 - 266 - 200 - 289
3 0 -245 25 - 83 20 - 23 27 (no RP available)
4 A -132 17 - 69 10 + 29 10 - 600 - 520 - 468 0 - 388 - 439 - 359
6 A -211 13 - 400 - 781 - 181 1 - 570
II - 400 - 227 - 181 1 - 16
7a A -260 17 -1000 - 703 - 740 0 - 443
Ic -1050 - 694 - 790 0 - 434
b A -251 17 "c - 450 - 368 - 199 2 - 117
8 A -204 16 "c - 475 - 479 - 271 1 - 275

* All values for subjects S.B., G.L. and B.D. are for RPs recorded at the vertex. For subjects S.S. and C M . the relevant RPs were recorded only at the
contralateral prccentral area for the hand, as designated by subscript c, except for some vertex recordings noted by subscript v. Simultaneous values for v
and c are given for sessions 6 and 7a of subject C M . f SE = standard error for our estimate of the mean value.
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 631

to zero (actual stimulus time) or possibly delayed slightly. But the actual mean values
for S were usually negative rather than positive or delayed, except for subject B.D.,
and they differed for each subject and with each session. The value obtained for S in
a given session could be regarded as at least a partial measure of the way the subject
is handling those reporting factors that S and W series do have in common. As an
approximation, one may 'correct' W for the subject's 'bias' in reporting awareness
time by our methods, by subtracting S from W for that given session. The average of
all W values (about 200 ms) would be changed to about 150 ms by subtracting
the average of about 50 ms for all S values (see Table 2D).

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

TABLE 2. GRAND AVERAGES (MS) FOR ALL SERIES IN EACH COLUMN OF TABLE 1,
ACCORDING TO TYPE OF ASSOCIATED RP AND TO ORDER OF W AND M SERIES IN
A SESSION
Type of RP, c. B-A; i.e., (Onset RP)
for W B. Onset of RP (Onset RP) minus ( W-S),
series A. Awareness (in W series) minus ( W), using using
times
n W RP RP
np n RP RP
MN 9OX 90% MN 90%
I 12 -233 -1025 -784 -825 -522 6 -950 -585
II 20 -192 -535 -527 -343 -333 14 -366 -323
III 5 -183 -270 -517 -87 -334 3 -118 -409

D. Awareness times

W M S1

n X n X n X
For all series 37 -204 20 -86 22 -47

In sessions when
W series done before M series 10 -191 10 -92 10 -41

In sessions when
M series done before W series 10 -240 10 -80 10 -53

Mean values for M series were also mostly negative (except for subject B.D.),
averaging about 85 ms for all mean Ms (Table 2D). M was also slightly negative to
S in almost every individual study session (see Table 1, column A); so that even if the
average of M values ( 86 ms) are 'corrected' by subtracting the average of S values
( 47 ms), a small average net M of about 40 ms still remains. (Even for subject
B.D., subtracting the average of his mean Ss, + 83 ms, from the average of mean Ms
of +51 ms in the same 4 sessions, produces a net 'corrected' average M of about
30 ms.) This produces the unexpected result that reported time of awareness of
'actually moving' generally preceded the activation of the muscle at EMG zero time!
(See Discussion.)
It is important to note that mean W values were consistently negative to mean M
values in the respective session for each subject (Table 1A), in spite of the frequently
negative values for M. The average of all mean Ws (about 200 ms) indicated that
632 BENJAMIN LIBET AND OTHERS

awareness of wanting to move preceded average awareness of actually moving

(about 85 ms) by more than 100 ms. When only those W values obtained in the
same 20 sessions with M values are included, the average of mean Ws was 216
instead of 204 ms (Table 2D). Mean Ws obtained in sessions when an M series was
carried out before the W series appeared to be significantly more negative than Ws
obtained when a W series was carried out before an M series {see separate averages
in Table 2D). A Wilcoxon test for this ranking order gave a one-sided P = 0.038, and
a two-sided P = 0.076. (This ranking was not related to the use of (A) vs (O) mode of
recall in the session.) However, the actual differences of about 50 ms between the
two sets of Ws are relatively unimportant when comparing W times to onset times of

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

the corresponding readiness potentials {see Section III, below).

II. Onset Times of Readiness-potentials {RP)

RPs associated with the freely voluntary, self-initiated movements employed in
this study have been described (Libet et al., 1982). They can be categorized into two
or three types, based on their form and the time of the main negative (MN) shift {see
Tables 1B and 2B). Type II (and III) RPs are obtained when all 40 self-initiated
movements in the averaged series are reported by the subject to have originated
'spontaneously' and 'capriciously', with no recollections of preplanning experiences
for any of the 40 events in the series. Additional experiences of a 'preplanning' phase
are associated with type I RPs (Libet et al., 1982). No significant association could
be detected between mode of recall for W (that is (A) or (O)) and type of RP
obtained.
Onset times of RPs listed in Table 1B are for RPs recorded in the same series of 40
self-initiated movements for which the reports of W times are given, in each
respective session for each subject. This simultaneity, for RP and W observations, is
important because there can be considerable variations of RP onsets in different
series even in the same session (Libet et al., 1982). The actual RPs for each W series
listed in Table 1 for the Group 2 subjects (S.B., G.L. and B.D.) are presented in fig. 2.
RPs were also obtained with each M series in the session, but onset times for these
are not listed in Table 1. Onset times for RPs in M series were actually, on average,
similar to those for RPs in the W series {see also Libet et al., 1982).
Two values for onset time are given for each RP (W series) in Table 1B. (1) Onset
time of the main negative (MN) shift was determined by 'eye-ball inspection',
checked independently by a second investigator. (2) Onset time was also com-
puted for the point at which 90 per cent of the area under the RP tracing preceded
EMG zero time.
Onset time based on RP area was determined as follows. On an enlarged projected image, the area
under the RP was measured for each interval of 50 ms, starting from EMG zero time and progressing to
successive intervals in the negative (pre-EMG) direction until 600 ms; between 600 and 1400 ms,
areas for 100 ms intervals were measured. Within each time interval, any areas below the baseline were
subtracted from those above. In estimating total area, however, it was considered advisable to exclude
any early brief shifts of potential that did not continue progressively into the main RP, as some of these
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 633

S.B. G.L. B.D.

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

5a

5b

6a

6b
500 ms
FIG. 2. Readiness-potentials (RPs) recorded at the vertex and averaged for each (W) series of 40 trials for subjects
S.B., G.L. and B.D. Each RP corresponds to the respective (W) series as listed by session number in Table 1. The
solid vertical line indicates the EMG zero time, marking the end of the RP. Dashed horizontal lines represent the d.c.
baseline drift for the 2 s of that tracing, as estimated from the total voltage compensation for shift in d.c. level during
the total time between beginning and end of that series of 40 trials.

could have been artefactual in nature. Therefore, the rule was adopted that any 200 ms segment having
a total area < 4 mm 2 (actually equivalent to 50 /tV-ms) was to be regarded as zero, and that any and all
areas preceding that segment were also regarded as zero. (Making the rule even more stringent, by
reducing the excluding low level segment to 100 ms, rarely changed the results significantly and, when it
did, the final estimates of time of onset were very little different.) It was also recognized that
measurements of the beginning of the negative potential shift are subject to some possible error in
judging the d.c. baseline, especially in a somewhat noisy/bumpy tracing. Therefore, after arriving at a
total area for a given RP under the rule above, the time interval that included only 90 per cent of this
area was computed. The 90 per cent values also fit the range of differences between the independent
measurements of the areas by two different investigators.
634 BENJAMIN LIBET AND OTHERS

Averages of the onset times for RPMN and RP90% area, respectively, for the Ws
series in Table 1B are given in Table 2. The initially slower but progressive ramp-like
rise of type I RPs accounts for these onset times being more negative for RPMN than
for R P ^ area. On the other hand, in types II and III RPs some definitely
distinguishable negativity is often present even before the main (MN) shift. Such
negativities tend to have a relatively irregular, low, amplitude but there was no
reason to regard them as other than actual RP components in these self-initiated
acts (see Libet et al., 1982). Their inclusion in the measurements of total area makes
it possible for onset of R P ^ area to precede onset of RPMN in some cases. Averages
of the onsets for RP MN and R P , ^ were in fact not very dissimilar for type II RPs,

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

although individual values for the difference (RPMN RIV/J) were in a range between
- 207 and + 526 ms; but for type III RPs average R P ^ preceded RPMN by - 247 ms
(range of RPMN R P , ^ was between 115 and +556 ms).

///. Differences between Onset time ofRP and Time of Awareness of Wanting to Move
The data comparisons given in Table l c are central to the objective of this study;
they relate the time of appearance of the conscious intention to act, on the one hand,
to the time of onset of the cerebral processes before the act (as evidenced in the RP),
on the other. The difference between RP onset time and each W awareness-time is
given for each respective series of self-initiated voluntary acts. Differences are
presented when utilizing the W times as actually reported (W 'uncorrected'), giving
(onset RP) minus (W); or the W times 'corrected' by subtracting the reported mean
time for the S obtained in the same session, giving (onset RP) minus (W-S).
'Correcting' the W value by subtracting the S value of each subject's 'bias' in
reporting, did not qualitatively change the relation of RP onset-time to W; rather it
generally increased the difference by which onset ofRP precedes W (as 'corrected').
For subject B.D., his positive values for S have the oposite effect; but even for him,
the only qualitatively important change in the difference is introduced in session 1,
which had a large positive S (+135 ms).
It may be seen (Table lc) that, with few exceptions, onset ofRP occurred before
reported awareness time by substantial amounts of time. This was true irrespective
of which measure of RP-onset or of W is employed to obtain the difference. The sizes
and consistency of these differences, between onset of RP and W, indicate they are
highly significant. However, it is difficult to produce a rigorous quantitative value
for significance of the large differences between onset of RP and W. An SD
(standard deviation) for variability among individual RPs within each series of 40 is
not available, as only the average RP for the whole series could be meaningfully
recorded. Consequently, only the mean W value and the averaged RP obtained in a
given series can be compared for statistical purposes. Confidence in the significance
of the differences (in Table lc) is further raised by the fact that they were almost
invariably very large when compared to the SEs (standard errors) for the mean
values of W (Table 1A). In addition, each W series of 40 trials was examined for the
incidence of individual W values that may have deviated sufficiently from the mean
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 635

W so as to be negative to (precede), rather than positive to (follow) the onset of

average RP in that series {see Table 1 c). For 36 W series of 40 trials each, instances in
which individual W time preceded onset time of averaged RP numbered zero in 26
series and 1 to 3 in 8 series! (Of the remaining 2 series, in both of which RPs were
recorded at contralateral precentral sites where RP onsets are often less steep than at
the vertex, 1 had 4 and 1 had 17 instances. The latter large value holds only in
relation to onset of MN in a type III RP but not if onset R P ^ is used.)
The SE of each mean value for W, as given in Table 1A, is more meaningful in relation to difference of
(onset of averaged RP)-(mean W), than is the SD for the distribution of the individual W values in each
series of 40 trials. (The SD for each series of 40 W values may be calculated using the respective SE,

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

given in Table 1A; SD = V40 x [SE].) If the individual RPs were available, the difference between each
RP onset and the respective individual W value could be determined for each event, and a meaningful
evaluation of variability for such individual differences could be made. However, in the absence of
individual RP values, it is reasonable to assume that the W and RP-onset for each individual act are
probably related in a dependent and positively correlated manner, in which onset of each RP tends to
be negative to (before) each W. This condition seems likely because (1) W and RP are features of the
same underlying process, (2) observed differences between averaged RP and mean W were consistently
large, and (3) there was a nearly complete absence of individual W values that were negative to the
averaged RP {see above).

For reporting each W time, subjects were asked to note the earliest awareness of
the specific urge or intention to act which might occur prior to the act. All subjects
reported that they could distinguish readily between this awareness and any
experience of 'pre-planning' that sometimes occurred in acts associated with type I
RPs (Libet et al., 1982). Awarenesses of 'preplanning' were completely absent in
series associated with type II (or III) RPs, in which all 40 self-initiated movements
were 'spontaneous' in origin. Therefore, it is useful to consider the values for type II
(and III) RP series separately from those for type I RP series, as summarized in Table
2A-C. AS might be expected, series with type I RPs generally exhibit an earlier onset
of RP, relative to W, than do those with type II (or III), especially for RPMN onsets.
However, even for the series of 'spontaneous' acts with types II and III RPs, onsets
of RPs generally preceded W by substantial amounts. The average of the differences
[(onset RPMN) minus (W)] for the 20 series with type II RPs was - 343 ms (Table 2c).
The relatively few sessions in which these differences were possibly not significant
occurred mainly in association with type III RPs, for which onsets of RPMN averaged
only 270 ms. For the 5 sessions with type III RPs, the average difference (onset
RPMN W) was 87 ms, although when R P ^ is used, the difference increased
to - 334 ms.

DISCUSSION

It is clear that neuronal processes that precede a self-initiated voluntary action, as

reflected in the readiness-potential, generally begin substantially before the reported
appearance of conscious intention to perform that specific act. This temporal
difference of several hundreds of milliseconds appeared fairly consistently regardless
636 BENJAMIN LIBET AND OTHERS

of which of the available criteria for onset of RP or for the time of awareness are
adopted. Series with type II RPs are of especial interest as all of the 40 self-initiated
acts arise spontaneously; on this and other evidence, the main negative (MN) shift
with average onset about 550 ms was postulated to reflect the cerebral volitional
process uniquely involved in initiating a freely voluntary, fully endogenous act
(Libet et al., 1982). Even for such series, with type II RPs, onset of RP preceded W by
about 350 ms on the average. In series with type I RPs the earlier MN shift (average
onset about 1025 ms) appears to reflect a more general preparation or intention to
act that can be either endogenous or cued externally; it is not necessarily associated
with freedom of choosing when to act (Libet et al., 1982). However, actual

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

experiences of 'preplanning' were reported for only a minority of self-initiated acts
in series with type I RPs. Consequently, the much larger differences between onset of
type I RP and W, on the average as much as 800 ms, may also reflect advance
cerebral preparation that is generally accomplished before conscious intentionality
arises. Only in the case of a small number of series with type III RPs was the
difference between RP-onset and W less negative than 100 ms, when onset of MN
shift (average 270 ms) is adopted. But if start of 90 per cent of RP area (average
517 ms) rather than MN shift is taken as the criterion for onset of RP, even
these type III RP's precede W by an average of more than 300 ms (Table 2c). In
series with type III RPs, all self-initiated acts were also spontaneous, as in type II
(Libet etai, 1982).
The validity of the RP as an indicator of cerebral activity had already been
established (Kornhuber and Deecke, 1965; Deecke et al., 1976; Pieper et al., 1980).
Actually, the onset of RP provides only a minimum timing for initiation of cerebral
activity, as the recorded RP probably represents neuronal activity in a limited
portion of the brain, possibly that of the supplementary motor area in the mesial
neocortex (Deecke and Kornhuber, 1978; Eccles, 1982; Libet et al., 1982). It is
possible that cerebral activity is initiated at times earlier than the onset of the
recorded RP in some other regions (see Groll-Knapp et al., 1977). For the present
issue, the requirement of averaging 40 individual RPs for each recordable RP might
raise the question whether this obscures an actually randomized group of widely
different onset times. Even in such a case, however, there are several considerations
that work mathematically against an interpretation that W times are consistently
equal to or precede onset times of individual RPs. Some of these have been discussed
above (Section III, Results). An example might assume that the averaged RPs are
contaminated by a small number of unusual individual RPs having very early
onsets, compared to onsets of most RPs in each series of 40. However, if this were
true, we would expect that such unusual RPs should not appear in an appreciable
number of the 40-trial series, causing the order of times for W and onset of averaged
RP to be reversed; but, in fact, the onset of averaged RP preceded mean W in all of
the 37 W series (Table lc, and see above). Furthermore, in type II RPs the main
negative shift (MN) has a relatively abrupt onset and rises rapidly; any small
number of individual RPs with unusually early onsets could not appreciably affect
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 637

the onset time of RPMN, as measured in type II RPs. Yet, onset of RPMN preceded
mean W, usually by large time intervals, in all the 20 W series that exhibited type II
RPs. To circumvent all these points of evidence one would have to introduce
unsupported assumptions that there exist specially biased distributions of RPs.
As another possibility, it might be proposed that neural activities, represented by individual RPs
with randomly variable amplitudes and onset times, must achieve some threshold, whether integrative
or other, before the brain 'decides' to act voluntarily; for this one might apply a kind of'random-walk'
model for sequential decision making (see Audley, 1973). If applied to our case, a delayed appearance
of a 'neural threshold' might coincide in time with the W time, and thus nullify any apparent
discrepancy between time of cerebral decision and time of conscious intention to act. But a consistent

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

bias or change in the random-walk distribution of the variable neural functions would have to be
initiated at the onset of averaged RP, preceding the achievement of threshold for the decision. This
would be analogous to the required initiation and presence of a sensory signal to produce the
distribution of sensory random variables that may lead to a yes-no decision in that random-walk model
(Green and Luce, 1973). Therefore, such a proposal could, at best, only separate the cerebral initiating
process into two stages. An earlier stage, which would precede a final decision stage, would start with
some initiating endogenous trigger. Such a model would not fundamentally affect our conclusion, that
cerebral activity initiates the voluntary act before reportable conscious intention appears.

The Criteria for Time of Conscious Intention to Act

The reliability and validity of these operational criteria are of course crucial to the
issue of the temporal order of cerebral processes vs conscious intention. The
reliability of the subjects' reports of'clock-position' for the revolving CRO spot at
the time of awareness appears to be fully adequate (see discussion of SEs of mean W
values and incidence of individual W values, relative to onset of averaged RP, in
Section III of Results).
Consideration of validity of our criteria begins with the premise that the
subjective event in question is only introspectively accessible to the subject himself,
and that this requires a report by the subject (see Libet, 1965, 1966; Nagel, 1974;
Creutzfeldt and Rager, 1978). Any behavioural response that is not a direct function
of such a report could not be used as a primary indicator of the subjective event
(Libet, 1981a, b), although it might be found to be associated with the subjective
event as studied by suitably valid reports. Acceptance of this premise, and of our
specific operational precedure for the required introspective report, introduces
several issues that may affect the validity of the reported W time. In particular,
factors that may affect the transmission between the subject's introspective
experience and his verbal report must be considered.
(1) Simultaneity of judgements. Our method requires the subject to observe
simultaneously, for later report, the appearance of a mental event (conscious urge to
move) and the visual clock-position of the revolving spot of light at that time.
Reports of simultaneous events have long been known to be subject to potential
errors, depending on the circumstances (differential attentiveness, in the 'prior
entry' phenomenon) and on the individual subject (see Boring, 1957; Efron, 1973;
Sternberg and Knoll, 1973). Our S experimental series, in which subjects reported
awareness times for skin stimuli, were designed to serve as controls for potential
638 BENJAMIN LIBET AND OTHERS

errors in such 'simultaneity' as well as for other individual biases and errors in the
entire reporting procedure. Procedures and requirements for subject's attentiveness,
observations and later recall, of clock-positions of the revolving CRO spot at the
time of awareness of a randomly appearing skin sensation, were the same as in the W
and M series. But in the S series the actual time of the stimulus was later known to
the investigator, and the error in the subject's reports could be determined
objectively. The bias or error found in the S series did not qualitatively alter the
difference between onset of RP and W, as determined in a given session for the same
subject; in fact, they generally enlarged the differences (Table 2c).
(2) Timing of an endogenous mental event. The subject's reported time for

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

spontaneously arising awareness of the urge/intention to move cannot be directly or
objectively validated in the manner possible for skin stimuli in the S series. The
subject's report constitutes the primary evidence of his introspective experience. No
other independent measure of such subjective timing is available, although some of
the other available evidence can affect confidence in the validity of the observed
timings; this is summarized below in point (3).
It is of course possible to conceive or postulate conditions that might introduce discrepancies
between the actual and the reported initial times of such an awareness. For example, what if it were
possible to judge accurately only the end of the mental event, the conscious urge to move; the actual
time of its onset, in relation to a perceived clock-time, would then be unknown and in doubt. In relation
to such a suggestion we note,first,that each subject was instructed to 'watch for' and report the earliest
appearance of the awareness in question, and subjects did not raise any difficulties about doing this.
Secondly, perceptual timings of onset and offset found for a peripherally induced sensation, at least, do
not support the suggestion that only the end is accurately judged. Using a method of cross-modality
simultaneity judgements, Efron (1973) found that (a) there was no difficulty in distinguishing onsets
from offsets; (b) the perceptual onset latency was constant regardless of large changes in duration of the
stimulus; whereas (c) perceptual offset latency could in fact vary when stimulus durations were
shortened to < 150 ms, with the change in offset latency probably originating, in part, in the
peripheral sensory structures. Somewhat related findings on perceptual onset latencies are available for
a cerebrally induced mental event, elicited by a stimulus in the somatosensory system in man (Libet et
al., 1979). The mean differences between perceptual timings, for a brief skin stimulus relative to onset
of a medial lemniscus stimulus lasting 200 ms, were only within a few tens of milliseconds. One might
suggest that timing of an endogenous mental event, the spontaneously arising conscious intention to
act, may be more difficult subjectively to pinpoint with accuracy than the timing even for the sensation
elicited by an intracerebral stimulus in the medial lemniscus. The individual W values reported by our
subjects in a given series of 40 events did show variability; but the difference between (onset of RP) and
(mean W), in Table lc, was consistently and considerably greater than the SE of mean W for the
respective series. Actually, the converse possibility, of W times reported earlier than the actual time of
awareness of the urge to move, presented a more real difficulty. A 'preawareness' that one is preparing to
perform the voluntary act, sometime within the next second or so, does in fact accompany at least some
of the events in those series that produce a type I RP, as noted above (Libet et al., 1982). If such a
preawareness were to have affected the report, it would mean that the reported W times were more
negative, earlier, than they should have been; the difference between onset of RP and 'real' W should
then be even greater than indicated by our results. However, in series giving type II RPs, all of the self-
initiated acts were described as 'spontaneous'; the subjects reported that each urge or wish to act
appeared suddenly 'out of nowhere', with no specific preplanning or preawareness that it was about to
happen.
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 639

Finally, the possibility could be raised that an earlier nonrecallable phase of the
conscious urge exists, one that is not storable as a short-term memory. If it is
further assumed that the subject's report of W time requires short-term memory of
the mental event, then the reported time would apply only to a later, recallable phase
of awareness, given such assumptions. First, one should note that, to report W time,
the subject needed to recall only the clock-position of the revolving spot at the time
he became aware of the urge/intention to move, and not necessarily the conscious
mental event itself. (Actually, the latter was at least often recallable, as the subjects
were able to describe it, even in relation to experiences just preceding it during the
trial.) Secondly, the proposal of a nonrecallable initial phase of the conscious urge to

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

move is a hypothetical construct which, like some other potential uncertainties in
timing an endogenous mental event, is at present not directly testable. Thus,
although it cannot be definitively excluded, it also lacks experimental support.
(3) Additional evidence bearing on the validity of the reported timings. One way to
test and improve confidence in the validity of the reported timings lies in using
different but converging modes of observing and reporting, with each mode having
independent validity without further assumptions {see Garner, 1954, for discussion
of this approach in connection with a related issue). Two quite different modes were
employed for reporting the 'clock-positions' of the CRO spot at the time of
awareness, that is, the absolute (A) reading vs the order (O) relative to final spot
positions. Yet both modes produced values for W in the same range and were
essentially indistinguishable. (This also held for reported timings in the M and S
series.)
Subjects definitively distinguished the experience and time of awareness of
wanting to move (W) not only from those of a skin sensation (S) but also from
awareness of actually moving (M). Mean values for W times were consistently
negative (by > 100 ms on average) to those for M times (Tables 1A, 2D). This was
true in spite of the unexpectedfindingthat mean M values were themselves generally
negative to EMG zero time, although they were only slightly negative to S values in
which no movement was involved. M has some features of an endogenous mental
event, rather than simply of a sensation elicited by input from peripheral sensory
sources (see below, for discussion of M preceding the EMG). On the other hand,
onset-times for type I or II RPs in series asking for W reports were generally similar
to those asking for M reports, or even to those in series when no reports of awareness
were requested (see also Libet et al., 1982). This indicates that the somewhat
different mental sets associated with each kind of reporting (or absence of reporting)
did not affect the RP side of the time differential with respect to W.
It might be argued that subjects' reports of W times could be distorted by their
awareness of the time of the actual movement; this might induce them to report W
times that are later than the actual time of appearance of the conscious intention to
move. But the subjects confirmed that for W reports they concentrated on noting
their earliest awareness of any urge/intention to move. They further stated that their
mental set for 'observing' W time was also different from that for M time.
640 BENJAMIN LIBET AND OTHERS

Furthermore, the available evidence indicates that the subjects' experience with
attending to the awareness of actually moving (M) may have induced them to report
W times that were somewhat more, not less, negative relative to EMG zero time {see
Results, end of Section I). The mean W values were on the average about 50 ms more
negative when, in a given session, the W series was performed after instead of before
an M series {see Table 2D, etc.).
Awareness of'actually moving' (M) preceded the EMG. Mean M values were generally negative to
EMG zero time for most subjects, and consistently though slightly negative (average about 40 ms)
relative to S values for all subjects. Timing of M so as to precede the activation of muscle contraction
indicates that M was not reflecting awareness of proprioceptive sensory impulses elicited by the

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

movement. It suggests the possibility that M reflected an awareness associated more immediately with
initiation of efferent cerebral output for the movement. Components that follow the main negative
RP shift are recordable just prior to movement, including a negative 'motor potential' that begins
about 50 ms before the EMG {see Deecke et al., 1976; Shibasaki et al., 1980).

Unconscious and Conscious Initiation or Control of Voluntary Acts

Since onset of RP regularly begins at least several hundreds of milliseconds before
the appearance of a reportable time for awareness of any subjective intention or
wish to act, it would appear that some neuronal activity associated with the eventual
performance of the act has started well before any (recallable) conscious initiation or
intervention could be possible. Put another way, the brain evidently 'decides' to
initiate or, at the least, prepare to initiate the act at a time before there is any
reportable subjective awareness that such a decision has taken place. It is concluded
that cerebral initiation even of a spontaneous voluntary act, of the kind studied here,
can and usually does begin unconsciously. The term 'unconscious' refers here simply
to all processes that are not expressed as a conscious experience; this may include
and does not distinguish among preconscious, subconscious or other possible
nonreportable unconscious processes.
A general hypothesis had already been proposed that some substantial time
period of appropriate cerebral activity may be required for eliciting all specific
conscious experiences (Libet, 1965). This developed out of experimentally based
findings that cortical activities must persist for up to 500 ms or more before
'neuronal adequacy' for a conscious sensory experience is achieved {see Libet, 1966,
1973, 1981a; Libet et al., 1972). In that hypothesis, those cerebral activities that did
not persist sufficiently would remain at unconscious levels. The present evidence
appears to provide support for that more general hypothesis. It suggests that a
similar substantial period of cerebral activity may also be required to achieve
'neuronal adequacy' for an experience of conscious intention or desire to perform a
voluntary act.
The present evidence for the unconscious initiation of a voluntary act of course
applies to one very limited form of such acts. However, the simple voluntary motor
act studied here has in fact often been regarded as an incontrovertible and ideal
example of a fully endogenous and 'freely voluntary' act. The absence of any larger
meaning in the simple quick flexion of hand or fingers, and the possibility of
 CEREBRAL AND CONSCIOUS TIMES OF VOLITION 641

performing it with capriciously whimsical timings, appear to exclude external

psychological or other factors as controlling agents. It thus invites the extrapolation
that other relatively 'spontaneous' voluntary acts, performed without conscious
deliberation or planning, may also be initiated by cerebral activities proceeding
unconsciously.
These considerations would appear to introduce certain constraints on the
potential of the individual for exerting conscious initiation and control over his
voluntary acts. However, accepting our conclusion that spontaneous voluntary acts
can be initiated unconsciously, there would remain at least two types of conditions
in which conscious control could be operative. (1) There could be a conscious 'veto'

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

that aborts the performance even of the type of 'spontaneous' self-initiated act
under study here. This remains possible because reportable conscious intention,
even though it appeared distinctly later than onset of RP, did appear a substantial
time (about 150 to 200 ms) before the beginning of the movement as signalled by the
EMG. Even in our present experiments, subjects have reported that some recallable
conscious urges to act were 'aborted' or inhibited before any actual movement
occurred; in such cases the subject simply waited for another urge to appear which,
when consummated, constituted the actual event whose RP was recorded (Libet et
al, 1982). (2) In those voluntary actions that are not 'spontaneous' and quickly
performed, that is, in those in which conscious deliberation (of whether to act or of
what alternative choice of action to take) precedes the act, the possibilities for
conscious initiation and control would not be excluded by the present evidence.

ACKNOWLEDGEMENTS
We are indebted to an anonymous editorial reviewer of the paper for helpful comments. This work
was supported in part by the Research Support Program of the Mount Zion Hospital and Medical
Center, San Francisco.

REFERENCES
AUDLEY R J (1973) Some observations on theories of choice reaction time: tutorial review. In: A Mention
and Performance IV. Edited by S. Kornblum. New York and London: Academic Press.
BORING E G (1957) A History of Experimental Psychology. Second edition. New York: Appleton-
Century-Crofts, pp. 146-147.
CHURCH J D, COBB E B (1971) Nonparametric estimation of the mean using quantal response data.
Annals of Institute of Statistical Mathematics, 23, 105-117.
CREUTZFELDT O D, RAGER G (1978) Brain mechanisms and the phenomenology of conscious
experience. In: Cerebral Correlates of Conscious Experience. Edited by P. A. Buser and A.
Rougeul-Buser. Amsterdam: Elsevier/North Holland Publishing Company, pp. 311-318.
DEECKE L, GROZINGER B, KORNHUBER H H (1976) Voluntary finger movement in man: cerebral
potentials and theory. Biological Cybernetics, 23, 99-119.
DEECKE L, KORNHUBER H H (1978) An electrical sign of participation of the mesial 'supplementary'
motor cortex in human voluntary finger movement. Brain Research, Amsterdam, 159, 473-476.
ECCLES J C (1982) The initiation of voluntary movements by the supplementary motor area. Archivfiir
Psychiatrie und Nervenkrankheiten, 231, 423-441.
642 BENJAMIN LIBET AND OTHERS

EFRON R (1973) An invariant characteristic of perceptual systems in the time domain. In: Attention and
Performance IV. Edited by S. Kornblum. New York and London: Academic Press, pp. 713-736.
GARNER W R (1954) Context effects and the validity of loudness scales. Journal of Experimental
Psychology, 48, 218-224.
GILDEN L, VAUGHAN H G JR, COSTA L D (1966) Summated human EEG potentials with voluntary
movement. Electroencephalography and Clinical Neurophysiology, 20, 433-438.
GREEN D M, LUCE R D (1973) Speed-accuracy trade off in auditory detection. In: Attention and
Performance IV. Edited by S. Kornblum. New York and London: Academic Press, pp. 547-569.
GROLL-KNAPP E, GANGLBERGER J A, HAIDER M (1977) Voluntary movement-related slow potentials
in cortex and thalamus in man. In: Attention, Voluntary Contraction and Event-Related Cerebral
Potentials. Progress in Clinical Neurophysiology, Volume 1. Edited by J. E. Desmedt. Basel:
Karger.

Downloaded from http://brain.oxfordjournals.org/ at University of Plymouth on January 13, 2012

KORNHUBER H H, DEECKE L (1965) Hirnpotentialanderungen bei Willkiirbewegungen und passiven
Bewegungen des Menschen: Bereitschaftspotential und reafferente Potentiale. Pfliigers Archivfur
Gesamte Physiologie, 284, 1 -17.
LIBET B (1965) Cortical activation in conscious and unconscious experience. Perspectives in Biology
and Medicine, 9, 77-86.
LIBET B (1966) Brain stimulation and the threshold of conscious experience. In: Brain and Conscious
Experience. Edited by J. C. Eccles. Berlin: Springer-Verlag, pp. 165-181.
LIBET B (1973) Electrical stimulation of cortex in human subjects, and conscious sensory aspects. In:
Handbook of Sensory Physiology. Edited by A. Iggo. Heidelberg: Springer-Verlag, Volume 2,
pp. 743-790.
LIBET B (1981a) The experimental evidence for subjective referral of a sensory experience backwards in
time: reply to P. S. Churchland. Philosophy of Science, 48, 182-197.
LIBET B (19816) ERPs and conscious awareness; neurons and glia as generators. In: Electrophysio-
logical Approaches to Human Cognitive Processing. NRP Bulletin, Volume 20. Edited by R.
Galambos and S. A. Hillyard. Cambridge, Mass. The MIT Press Journals, pp. 171-175,226-227.
LIBET B, ALBERTS W W, WRIGHT E W, FEINSTEIN B (1972) Cortical and thalamic activation in
conscious sensory experience. In: Neurophysiology Studied in Man. Edited by G. G. Somjen.
Amsterdam: Excerpta Medica, pp. 157-168.
LIBET B, WRIGHT E W JR, FEINSTEIN B, PEARL D K (1979) Subjective referral of the timing for a
conscious sensory experience. A functional role for the somatosensory specific projection system
in man. Brain, 102, 193-224.
LIBET B, WRIGHT E W JR, GLEASON C A (1982) Readiness-potentials preceding unrestricted
'spontaneous' vs pre-planned voluntary acts. Electroencephalography and Clinical Neurophysio-
logy, 54, 322-335.
NAGEL T (1974) What is it like to be a bat? Philosophical Review, 83, 435-450.
PIEPER C F, GOLDRING S, JENNY A B, MCMAHON J P (1980) Comparative study of cerebral cortical
potentials associated with voluntary movements in monkey and man. Electroencephalography and
Clinical Neurophysiology, 48, 266-292.
SHIBASAKI H, BARRETT G, HALLIDAY E, HALLIDAY A M (1980) Components of the movement-related
cortical potential and their scalp topography. Electroencephalography and Clinical Neurophysio-
logy, 49, 213-226.
STERNBERG S, KNOLL R L (1973) The perception of temporal order: fundamental issues and a general
model. In: Attention and Performance IV. Edited by S. Kornblum. New York and London:
Academic Press, pp. 629-685.
(Received July 20,1982. Revised December 14, 1982)